Human immunodeficiency virus–related primary central nervous system lymphoma

Factors influencing survival in 111 patients

Authors

  • Mark E. Newell M.B.B.S., M.P.H.,

    1. National Centre in HIV Epidemiology and Clinical Research, University of New South Wales, Sydney, Australia
    Search for more papers by this author
  • Jennifer F. Hoy M.B.B.S.,

    1. Infectious Diseases Unit, Alfred Hospital, Melbourne, Australia
    2. Department of Medicine, Monash University, Melbourne, Australia
    Search for more papers by this author
  • Stephen G. Cooper M.B.B.S.,

    1. Department of Radiation Oncology, St. Vincent's Hospital, Sydney, Australia
    Search for more papers by this author
  • Bernadette DeGraaff B.Sc.,

    1. Infectious Diseases Unit, Alfred Hospital, Melbourne, Australia
    2. Department of Medicine, Monash University, Melbourne, Australia
    Search for more papers by this author
  • Andrew E. Grulich Ph.D.,

    1. National Centre in HIV Epidemiology and Clinical Research, University of New South Wales, Sydney, Australia
    Search for more papers by this author
  • Melissa Bryant B.Sc.,

    1. Infectious Diseases Unit, Alfred Hospital, Melbourne, Australia
    2. Department of Medicine, Monash University, Melbourne, Australia
    Search for more papers by this author
  • Jeremy L. Millar M.B.B.S.,

    1. William Buckland Radiation Therapy Centre, Alfred Hospital, Melbourne, Australia
    Search for more papers by this author
  • Bruce J. Brew M.B.B.S., M.D.,

    1. National Centre in HIV Epidemiology and Clinical Research, University of New South Wales, Sydney, Australia
    Search for more papers by this author
  • David I. Quinn M.B.B.S., Ph.D.

    Corresponding author
    1. Department of Clinical Pharmacology and Toxicology, University of New South Wales, Sydney, Australia
    Current affiliation:
    1. Division of Medical Oncology, Keck School of Medicine, University of Southern California, Los Angeles, California
    • Division of Medical Oncology, Keck School of Medicine, University of Southern California, 1441 Eastlake Avenue, Suite 3453, Los Angeles, CA 90033
    Search for more papers by this author
    • Fax: (323) 865-0061


Abstract

BACKGROUND

The current study evaluated factors influencing survival in patients diagnosed with human immunodeficiency virus (HIV)-related primary central nervous system lymphoma (PCNSL), with a focus on the effects of therapeutic radiotherapy (RT) and highly active antiretroviral therapy (HAART).

METHODS

A retrospective chart review of patients with a diagnosis of HIV-related PCNSL at one of five university hospitals between 1987 and 1998 was performed. Clinical details including antiretroviral agent use, brain imaging scan results, RT use, and survival outcomes were recorded.

RESULTS

One hundred eleven patients with HIV-related PCNSL were identified. The annual incidence decreased significantly between 1992 and 1995 and between 1996 and 1998 (P = 0.04). The median survival period was 50 days (mean, 109 days; range, 4–991 days), with improved survival for patients diagnosed after 1993. Patients treated with two or more antiretroviral agents had improved survival (P = 0.01), as did patients who received RT (P < 0.0001). For patients who received RT, completion of the prescribed course and treatment to ≥ 30 Gray (Gy) independently predicted a more favorable outcome. RT used in conjunction with antiretroviral therapy involving two or more agents had an additive positive effect on survival. For patients who did not receive RT, poor performance status and encephalopathy predicted a shorter survival duration.

CONCLUSIONS

The results of the current study suggest that HAART and treatment with RT to ≥ 30 Gy improve survival for patients with HIV-related PCNSL. This combination of therapies may provide a standard of care as the basis for further trials of chemotherapy, novel adjunctive treatment, and quality of life assessment. Cancer 2004. © 2004 American Cancer Society.

Before the onset of the human immunodeficiency virus (HIV) epidemic, primary central nervous system lymphoma (PCNSL) was a rare malignancy, accounting for < 1% of all primary intracranial tumors and only 1–2% of all non-Hodgkin lymphomas (NHL).1, 2 After the onset of the HIV epidemic in the early 1980s, the number of reported cases of PCNSL increased. Among HIV-infected patients, PCNSL is the second most common cerebral mass lesion after toxoplasmosis.3 It is still uncommon as an initial manifestation of acquired immunodeficiency syndrome (AIDS). For example, < 1% of HIV-infected patients are likely to develop PCNSL as the primary AIDS-defining illness. However, it is more common after the onset of AIDS, occurring in 2–13% of patients with a previous AIDS diagnosis.4 PCNSL tends to develop in patients with more severe immunodeficiency compared with patients with systemic NHL, with the median CD4-positive cell counts at diagnosis being 10 cells/mm3 and 189 cells/mm3, respectively, in these patients.5 Greater than 95% of AIDS-related PCNSLs are of B-cell origin and are associated with Epstein-Barr virus coinfection. The large cell immunoblastic pattern is the most common histopathologic pattern observed.4, 6

The median survival of patients with HIV-related PCNSL is poor. Several studies report a median survival of 3 months for those who receive radiotherapy (RT) and 1 month for those who do not receive treatment.7, 8 A number of cohort studies have suggested that RT, performance status, previous AIDS-defining illness, and CD4 count may predict outcome. It is unclear whether the survival difference in patients receiving RT is a primary beneficial effect that is attributable to the RT itself or to selection bias based on other prognostic factors that, in turn, influence the clinical decisions leading to the use of this therapeutic modality. One study suggests that HIV-infected patients with PCNSL who receive RT are less likely to die of PCNSL than of other mortal manifestations of AIDS, whereas patients not treated with RT are more likely to die of PCNSL.9 Moreover, the effect of highly active antiretroviral therapy (HAART) in patients who already have HIV-related PCNSL is not well defined; to our knowledge, a study in which survival improved in six patients who received HAART and a case report represent the only published experience in the medical literature.10, 11

The aims of the current study were to determine whether the incidence of HIV-related PCNSL has decreased with the introduction of combination antiretroviral therapy, to determine the prognostic factors associated with increased survival, to determine whether an independent survival benefit associated with RT was evident when other prognostic factors were controlled, and to attempt to identify a subgroup of patients who would be most likely to benefit from RT.

MATERIALS AND METHODS

With the approval of an ethics committee, case records of all patients diagnosed with PCNSL during the period 1987–1998 were audited from the medical records of five Australian university hospitals: Alfred Hospital (Melbourne), Fairfield Hospital (Melbourne), St. Vincent's Hospital (Sydney), Prince of Wales Hospital (Sydney), and Royal Perth Hospital (Perth). Each of these centers was a designated state referral center for the treatment of HIV disease and had on-site consultative and treatment services in neurosurgery, neurology, and radiation oncology. These centers treated 54% of the patients with AIDS in Australia during the study period. Patients with PCNSL were identified using a combination of hospital coding records, hospital-based HIV databases, the Australian national AIDS database, and databases from radiation oncology and anatomic pathology departments. Patients at the two Sydney hospitals were matched by linking a database of all AIDS-related NHL with the New South Wales Cancer Registry.12. Patients initially considered eligible for inclusion in the study but subsequently found ineligible were documented but not included in further analysis. Patients who were excluded had a diagnosis of CNS lymphoma but were found to have a previous or concurrent diagnosis of systemic NHL. Patients who presented with lymphoma confined to the leptomeninges but who did not have evidence of cerebral lesions after computed tomographic (CT) or magnetic resonance imaging (MRI) scans were excluded because of heterogeneity of clinical presentation and a significant chance of concurrent systemic NHL. This process yielded 111 evaluable patients with HIV infection with a diagnosis of PCNSL. Analyses were performed for these patients. Patients who completed RT planning were deemed to have commenced RT treatment for the purposes of the analysis.

Data Collection

Baseline data were collected, including gender, date of birth, date of AIDS diagnosis, list and date of other previous AIDS-defining illnesses if not PCNSL (including previously suspected toxoplasmosis), Eastern Cooperative Oncology Group (ECOG) performance status at admission, date of PCNSL diagnosis, presenting symptoms (classified as headache, seizure, ataxia, focal neurologic symptoms, and/or nonfocal neurologic symptoms [e.g., confusion, cognitive decline, and/or memory loss]), anatomic site(s) of PCNSL (classified as cerebral cortex, periventricular site, basal ganglia, brain stem, and/or cerebellum), antiretroviral therapy, and previous therapy for suspected toxoplasmosis. The date of diagnosis of PCNSL was defined as the date of the first abnormal CT or MRI scan showing a lesion considered to be PCNSL after nonresponse to a course of antitoxoplasma therapy (if applicable). Clinical data collected subsequent to a PCNSL diagnosis included the dates and details of brain biopsies; RT treatment details, including fields, prescribed and delivered doses, and fractionation; outcome; and survival. All patient data were stored in a confidential manner using a name code and date of birth in a password-protected computer database.

Statistical Methods

Statistical analyses were performed with Statview software (Version 4.5; Abacus Concepts, Berkeley, CA), using Cox regression univariate and multivariate analyses,13 Kaplan–Meier analysis,14 and log-rank tests, as well as the Fisher exact test as appropriate. Statistical significance in the current study was defined by P < 0.05. All reported P values are two sided. Univariate analysis was performed to identify baseline factors that had a significant impact on survival. Multivariate survival analysis was undertaken using the variables identified on univariate analysis as having prognostic significance. The effect of RT was tested by generating multivariate models that included as variables the administration of RT (RT treatment model), completion of RT (RT completion model), and receipt of a minimum radiation dose of 30 grays (Gy) (RT dose model) to determine whether any survival benefit persisted when other factors were controlled for. Given potential concerns and differing opinions in the medical literature regarding treatment effects in patients with a clinical but not a histologic diagnosis of PCNSL,15–20 a subgroup analysis of the 47 patients with biopsy-proven PCNSL was performed using the same three models.

RESULTS

Baseline Demographics

One hundred eleven patients with PCNSL were included in the database for analysis. One hundred eight patients were male, with a mean age of 38.0 years (median, 36.6 years; range, 22.0–63.2 years). Other baseline characteristics are shown in Tables 1 and 2. The original data abstraction identified an additional 14 patients presenting with lymphoma involving the CNS. However, a subsequent review of case notes revealed that these patients did not have PCNSL, but rather systemic NHL.

Table 1. Baseline Demographics for Patients with HIV-Related PCNSL (n = 111)
VariableEntire cohortRTP valueb
No (n = 57)Yesa (n = 54)
  • HIV: human immunodeficiency virus; PCNSL: primary central nervous system lymphoma; RT: radiotherapy; ECOG: Eastern Cooperative Oncology Group; AIDS: acquired immunodeficiency syndrome.

  • a

    Completed planning of treatment with radiotherapy.

  • b

    P value for difference between patients who received radiotherapy and patients who did not.

Gender    
 Male108   
 Female3   
Age (yrs)   0.09
 Median38.039.036.9 
 Mean36.639.735.3 
 Range22.0–63.222.0–58.423.9–63.2 
ECOG performance status   0.007
 Median2.52.72.3 
 Mean3.03.02.0 
 Range0–4.00–4.00–4.0 
No. of previous AIDS-defining illnesses   0.04
 Median1.82.01.52 
 Mean2.02.01.00 
 Range0–5.00–4.00–5.0 
Time from AIDS diagnosis to PCNSL diagnosis (days)   0.32
 Median487493480 
 Mean256295206 
 Range0–38280–38280–3232 
CD4 lymphocyte count (cells/mm3)   0.77
 Median37.425.550.1 
 Mean1111.010.5 
 Range1–7501–1501–750 
% not receiving antiretroviral therapy at PCNSL diagnosis34.233.435.20.84
Table 2. Univariate Analysis Results for Patients with HIV-Related PCNSL (n = 111)
VariableNo. of patients (%)Hazard ratio (95% CI)PaMedian survival (days)
  • HIV: human immunodeficiency virus; PCNSL: primary central nervous system lymphoma; CI: confidence interval; ARV: antiretroviral; CNS: central nervous system; ECOG: Eastern Cooperative Oncology Group; NHL: non-Hodgkin lymphoma; AIDS: acquired immunodeficiency syndrome; RT: radiotherapy; Gy: grays.

  • a

    P value for survival difference based on univariate analysis.

Age (yrs)    
 ≥ 40 42 (37.8)1.41 (0.95–2.09)0.09353
 < 40 69 (62.2)  50
Date of diagnosis    
 < 1/1/93 63 (56.8)1.56 (1.06–2.30)0.02649
 ≥ 1/1/93 48 (43.2)  60
CD4 count (cells/mm3)    
 ≤ 125104 (93.6)2.31 (1.05–5.09)0.03849
 > 125  7 (6.4)  110
ARV agents    
 0 or 1 86 (77.5)1.88 (1.16–3.02)0.009849
 2 or 3 25 (22.5)  59
Nonfocal CNS symptoms    
 Yes 57 (51.8)2.17 (1.47–3.20)< 0.000140
 No 53 (48.2)  98
Focal CNS symptoms    
 Yes 42 (38.2)0.91 (0.62–1.34)0.6350
 No 68 (61.8)  51
ECOG performance status    
 ≥ 3 67 (60.4)1.87 (1.26–2.76)0.001840
 ≤ 2 44 (39.6)  103
Biopsy-proven NHL    
 No 64 (58.2)2.00 (1.35–2.96)0.000540
 Yes 47 (41.8)  79
No. of previous AIDS-defining illnesses    
 ≥ 3 63 (56.8)1.47 (1.00–2.17)0.049748
 ≤ 2 48 (43.2)  65
Received RT    
 No 57 (51.4)3.05 (2.01–4.63)< 0.000151
 Yes 54 (48.6)  92
Completed RT    
 No 73 (65.8)3.50 (2.24–5.47)< 0.000140
 Yes 38 (34.2)  116
RT dose (Gy) received (n = 54)    
 < 30 18 (33.3)2.91 (1.59–5.33)< 0.000143
 ≥ 30 36 (66.7)  116

Incidence

The incidence of new cases of PCNSL, both as primary and subsequent AIDS-defining illnesses, increased markedly to a peak of 16 cases per year in 1992 and 1993, followed by a subsequent decline, with only two cases per year in 1997 and 1998 (Fig. 1). In assessing differences in incidence between time periods in the cohort, the overall trend in incidence was found to be significant (P = 0.03), as were the increases between 1987 and 1990 and between 1991 and 1993 (P = 0.03) and the downward trends between 1992 and 1995 and between 1996 and 1998 (P = 0.03). This latter change in incidence coincides with the onset of the use of HAART in Australia and a concurrent decrease in the incidence of AIDS dementia complex.21

Figure 1.

Incidence of human immunodeficiency virus–related primary central nervous system lymphoma per calendar year in the study cohort.

Presentation, Diagnosis, and Treatment of Primary Central Nervous System Lymphoma

All patients had neurologic symptoms and signs documented at presentation. For example, 61.3% of patients had focal neurologic symptoms, 51.4% of patients had nonfocal neurologic symptoms, 43.2% of patients had headache, 21.6% of patients had seizure, and 18% of patients had ataxia. All patients received a CT scan on presentation, with additional MRI for 18 patients (16.2%), 8 of whom had normal or nonspecific findings on CT scans. The anatomic sites of PCNSL at presentation were the cerebral cortex in 65% of patients, the periventricular region in 56% of patients, the basal ganglia in 33% of patients, the cerebellum in 7% of patients, and the brain stem in 4% of patients. Approximately one-half (48.6%) of the lesions were found in one region only, and the remaining lesions were found in two or more anatomic areas. Data on the number of lesions present on pretreatment CT or MRI scans were available for 90 patients. A mean of 2.8 lesions were present (median, 2 lesions; range, 1–16 lesions). No significant difference in the number of lesions at presentation was found when the periods before and after January 1993 were compared (P = 0.54). There was a significant association between the presence of periventricular PCNSL and nonfocal neurologic symptoms and signs, such as confusion, cognitive decline, and memory loss (P = 0.02). There also was an association between such symptoms and the presence of three or more lesions on imaging scans (P = 0.04). There were trends toward significance between seizures and the presence of PCNSL at periventricular (P = 0.06) and cerebral cortical sites (P = 0.1). Contemporaneous cerebrospinal fluid examination was performed for nine patients and yielded cytologically negative findings in all of them.

For 91% of patients, a presumptive diagnosis of cerebral toxoplasmosis based on CT or MRI findings was made, and these patients were treated with pyrimethamine and either sulfadiazine or clindamycin for a median treatment period of 15 days; dexamethasone therapy also was administered to the majority of these patients (88.3%). The rate of use of toxoplasmosis therapy before biopsy or before the start of treatment for lymphoma did not vary between the pre-1993 and post-1993 groups (P = 0.66). Subsequently, a diagnosis of PCNSL was made in 57.4% of patients based on a lack of imaging response to antitoxoplasma therapy. In the remaining 42.6% of patients, brain biopsy provided a histologic diagnosis. Eleven patients (10%) were judged to have concurrent PCNSL and toxoplasmosis. Data on histopathologic classification were available for 41 of the 47 patients who underwent biopsy. Nineteen patients had high-grade immunoblastic lymphoma, 21 patients had intermediate-grade diffuse large cell lymphoma, and 1 patient had small and cleaved cell lymphoma. Histopathologic classification did not influence survival (P = 0.48).

Approximately one-half of all patients (n = 54 [48.6%]) underwent planning for RT. Of these 54 patients, 36 (66.7%) completed the prescribed RT course. Of the patients who completed RT, 67% (n = 24) received a total dose ≥ 30 Gy. The total delivered dose range for patients completing the prescribed RT course was 15–56 Gy, all administered through opposed right and left lateral fields to the whole brain. Of the 54 patients receiving RT, 12 (22.2%) died of a CNS syndrome, presumably related to PCNSL, whereas 40 of 57 patients (61.4%) who did not receive RT died of a CNS syndrome (P < 0.0001). Of the 25 patients who received two or three antiretroviral drugs, 7 (28%) developed a new AIDS-related infection after the diagnosis of PCNSL, whereas 36 of 86 patients (42.8%) who received 0 or 1 antiretroviral drug developed such an infection (P = 0.008).

Survival

Univariate analysis

The median survival of the cohort was 50 days (mean, 109 days; range, 4–991 days). On univariate analysis (Table 2, Figs. 2 and 3), baseline parameters that had a significant positive impact on survival were diagnosis after January 1, 1993, a CD4-positive lymphocyte count ≥ 125 cells/mm3, previous antiretroviral therapy with two or more drugs, the absence of nonfocal CNS symptoms, an ECOG performance status of ≤ 2, biopsy-proven NHL, receipt of any RT, completion of RT, and a received RT dose of ≥ 30 Gy. Neuroanatomic site and number of lesions present had no significant impact on survival (P > 0.2; log-rank test). Patients who did not receive antiretroviral therapy before PCNSL diagnosis had an outcome similar to that of patients who received previous antiretroviral therapy (P = 0.35; log-rank test). Among patients who subsequently received more than two antiretroviral drugs, RT (P = 0.01), completion of prescribed RT (P = 0.003), RT to ≥ 30 Gy (P = 0.01), and biopsy confirmation of PCNSL diagnosis (P = 0.003) were associated with improved survival.

Figure 2.

Survival according to (A) radiotherapy (RT) use (n = 111; log-rank P < 0.0001), (B) completion of the prescribed RT course (n = 111; log-rank P < 0.0001), and (C) RT dose received (n = 54; log-rank P < 0.0001). Gy: grays.

Figure 3.

Survival according to ongoing antiretroviral therapy. (A) Entire cohort (n = 111; log-rank P = 0.008). (B) Subset of patients receiving radiotherapy (RT; n = 54; log-rank P = 0.07). (C) Subset of patients receiving a total RT dose ≥ 30 grays (n = 36; log-rank P = 0.02). ARVA: antiretroviral agent.

Multivariate analysis

Table 3 summarizes the multivariate analyses undertaken using the three RT models: the RT treatment model, the RT completion model, and the RT dose model (which is based on a total dose cutoff of 30 Gy, the dose considered the standard of care in most Australian and international centers7). Different cutoff points for RT doses received were explored on univariate analysis for their effects on survival: 20 Gy, P = 0.01; 30 Gy, P = 0.0006; and 40 Gy, P = 0.43. Increasing radiation dose was associated with improved survival as a continuous variable (P = 0.008). Additional significant independent predictors of improved survival in the RT treatment model were a CD4 lymphocyte count ≥ 125 cells/mm3 and the absence of nonfocal neurologic symptoms and signs. Using the RT completion model, only the absence of nonfocal neurologic symptoms and signs was found to be a significant predictor of improved survival in addition to RT dose ≥ 30 Gy. However, there was a nonsignificant association between improved survival and CD4 count ≥ 125 cells/mm3 (P = 0.08) and between improved survival and ongoing therapy with at least two antiretroviral drugs (P = 0.09). In the RT dose model, the significant predictors of improved survival that were used included ongoing combination antiretroviral therapy with at least two agents, the absence of nonfocal neurologic syndrome, and no more than two previous AIDS-defining illnesses, in addition to intent to treat with RT. On backward-stepwise regression modeling, the most predictive factors of improved survival were found to be receipt of RT, RT completion, and total RT dose ≥ 30 Gy, followed by (in decreasing order of effect on survival), ongoing therapy with two or more antiretroviral drugs, no more than two previous AIDS-defining illnesses, and the absence of nonfocal CNS symptoms and signs. Using multivariate analysis for the subgroup of patients who received treatment with two or more antiretroviral drugs, each of the RT-related variables was found to be independently predictive of an improved outcome, whereas no other variable achieved significance.

Table 3. Multivariate Analysis summary for Patients with HIV-Related PCNSL (n = 111)
VariableRT treatment model (n = 110)RT completion model (n = 110)RT dose model (n = 54)
Hazard ratio (95% CI)P valueHazard ratio (95% CI)P valueHazard ratio (95% CI)P value
  1. HIV: human immunodeficiency virus; PCNSL: primary central nervous system lymphoma; CI: confidence interval; ARV: antiretroviral; CNS: central nervous system; ECOG: Eastern Cooperative Oncology Group; NHL: non-Hodgkin lymphoma; AIDS: acquired immunodeficiency syndrome; RT: radiotherapy; Gy: grays.

Age (yrs)      
 ≥ 400.98 (0.62–1.57)0.960.88 (0.56–1.41)0.60.96 (0.45–2.01)0.91
 < 40      
Date of diagnosis      
 < 1/1/930.73 (0.41–1.30)0.290.80 (0.45–1.41)0.440.57 (0.21–1.58)0.28
 ≥ 1/1/93      
CD4 count (cells/mm3)      
 ≤ 1252.73 (1.15–6.50)0.022.24 (0.92–5.46)0.0762.65 (0.85–8.22)0.09
 > 125      
ARV agents      
 0 or 11.45 (0.81–2.56)0.211.66 (0.93–2.97)0.0872.31 (1.05–5.06)0.037
 2 or more      
Nonfocal CNS syndrome      
 Yes1.87 (1.20–2.92)0.0061.87 (1.18–2.94)0.0071.49 (0.72–3.09)0.28
 No      
ECOG performance status      
 ≥ 31.44 (0.93–2.23)0.101.38 (0.89–2.13)0.151.40 (0.70–2.80)0.34
 ≤ 2      
Biopsy-proven NHL      
 No1.35 (0.79–2.30)0.271.21 (0.71–2.05)0.481.07 (0.45–2.55)0.89
 Yes      
No. of previous AIDS-defining illnesses      
 ≥ 31.33 (0.86–2.06)0.21.29 (0.84–1.96)0.242.53 (1.30–4.95)0.007
 ≤ 2      
Received RT      
 No1.62 (0.96–2.71)0.069
 Yes      
Completed RT      
 None/not2.51 (1.44–4.36)0.001
 Completed      
RT dose (Gy) received (n = 54)      
 < 303.25 (1.47–7.21)0.004
 ≥ 30      

Additional analyses

In the subgroup analysis of patients who had biopsy-proven NHL (n = 47), univariate analysis revealed that the parameters that had a significant positive impact on survival included diagnosis after January 1, 1993, ongoing antiretroviral therapy with at least two agents, receipt of RT, completion of RT, and RT dose ≥ 30 Gy (Table 4). On multivariate analysis, ongoing antiretroviral therapy with at least two agents remained a significant predictor of survival along with RT in all three RT models. The significant correlation between poor survival and diagnosis before 1993 that was identified on univariate analysis was no longer significant when the number of antiretroviral agents was controlled for on multivariate analysis.

Table 4. Multivariate Analysis for the Subgroup of Patients with a Biopsy-Proven Diagnosis of PCNSL (n = 47)
VariableUnivariate analysisMultivariate analysis
RT treatmentRT completionRT dose
HR (95% CI)PHR (95% CI)PHR (95% CI)PHR (95% CI)P
  1. PCNSL: primary central nervous system lymphoma; HR: hazard ratio; CI: confidence interval; RT: radiotherapy; Gy: grays; ARV: antiretroviral.

Date of diagnosis        
 < 1/1/932.34 (1.12–4.10)0.020.77 (0.36–1.63)0.491.04 (0.48–2.25)0.931.13 (0.44–2.87)0.80
 ≥ 1/1/93        
ARV agents        
 ≥ 12.17 (1.13–4.15)0.021.90 (0.91–3.98)0.092.90 (1.30–6.49)0.012.81 (1.13–7.00)0.03
 ≤ 2        
Received RT        
 No2.81 (1.40–5.67)0.0042.54 (1.21–5.31)0.01
 Yes        
Completed RT        
 No2.87 (1.53–5.37)0.0013.74 (1.83–7.62)0.0003
 Yes        
RT dose (Gy)        
 < 302.75 (1.26–5.97)0.0084.40 (1.79–10.82)0.001
 ≥ 30        

Additional analysis was performed to identify potential prognostic factors associated with failure to complete RT. Significant factors included ECOG performance status ≥ 3 and age > 40 years (P = 0.04 and 0.05, respectively). In the subgroup of patients who did not receive any RT, the absence of nonfocal neurologic symptoms and ECOG performance status ≤ 2 were significant positive predictors of survival (P = 0.009 and 0.002 respectively; log-rank test). No other factors were predictive of survival in this subgroup.

DISCUSSION

The current study evaluated a large series of patients with HIV-related primary cerebral lymphoma using survival as the primary endpoint. The salient findings were that 1) combination antiretroviral therapy, especially with HAART, has markedly decreased the incidence of PCNSL in the study population; 2) ongoing treatment with combination antiretroviral therapy was associated with improved survival; 3) treatment with RT to a dose of 30 Gy was associated with improved survival; and 4) older patients (age ≥ 40 years) and patients with a poor performance status (ECOG performance status ≥ 3) were less likely to complete a prescribed course of RT.

Our reported overall median survival of 50 days, with significantly improved survival for patients receiving any RT (median, 92 days) compared with patients who did not receive RT (median, 38 days), is similar to the findings of other studies of HIV-related PCNSL. In a study of 55 patients performed early in the AIDS epidemic, the reported median survival was 119 days for patients who received RT, compared with only 27 days for patients who did not.8 In a cohort of 17 patients treated with varying doses of RT, the reported median survival was 72 days.22, 23 In a 10-year analysis of all patients with HIV-related PCNSL (n = 41) treated at the University of California–San Francisco (San Francisco, CA), the reported median survival durations were 3 months and 4.5 months for patients who received RT.24 A review of survival data for 174 patients in 32 series of AIDS-related PCNSL reported an overall mean survival of 2.6 months, 3 months for patients receiving RT and 0.9 months for patients not receiving RT.7 These individual studies, like ours, clearly were confounded by selection bias in the assignment of therapy, but they were performed before the advent of combination antiretroviral therapy.

To our knowledge, the current study is one of the first to demonstrate improved outcome for patients with HIV-related PCNSL treated with combination antiretroviral therapy. However, recent reports suggest that this effect may be relatively widespread, at least in the developed world.25, 26 In addition, there have been reports of clinical responses and disease remission in patients with PCNSL after the commencement of HAART with or without RT.10, 11 It is possible that the effect of antiretroviral therapy was influenced by the standard practice in the period before 1993 of ceasing administration of zidovudine (AZT) when RT was initiated due to concerns regarding myelosuppression and anemia during RT. However, the data in the current retrospective analysis do not allow evaluation of this effect.

The data regarding the recent decrease in incidence of HIV-related primary lymphoma in the current cohort are encouraging. It is likely that the sharp decrease in the incidence of cases following the rapid increase in the early 1990s is attributable to combination antiviral therapy, which became the standard of care in Australia after 1994 (Fig. 1). This sharp decrease in incidence has also been observed for other virally mediated opportunistic infections, such as Kaposi sarcoma and systemic NHL, in patients with HIV both in Australia27, 28 and elsewhere29, 30 and supports the hypothesis that improved immune status due to effective antiviral therapy resulted in the observed decreases in incidence. The sequential decreases in incidence in the current cohort between 1993 and 1994 and then between 1996 and 1997 may also reflect sequential changes in antiretroviral therapy during the study period. For example, the addition of didanosine to zidovudine occurred between 1992 and 1993, followed by the addition of protease inhibitors (HAART) beginning in 1995.21 The evidence to support this contention is limited, but these changes in therapy also coincide with downward inflection points in the overall incidence of AIDS in Australia.31–33 Similar patterns in the incidence of PCNSL and AIDS have been observed at a number of North American centers (Tupule A. Unpublished data),34 whereas other groups have reported only small decreases in PCNSL incidence over a similar time period.25, 26, 30, 35, 36 Further evidence in support of this sequenced effect on incidence is provided by data on 1255 patients diagnosed with AIDS between 1994 and 1997 in eight U.S. cities.37 That study demonstrated that increases in the intensity of antiretroviral therapy were associated with stepwise reductions in morbidity and mortality.37

Improved survival in patients selected for RT may be at least partially attributable to selection bias. Patients who received any RT had a better mean ECOG performance status (P = 0.007), fewer previous AIDS-defining illnesses (P = 0.04), and a nonsignificant trend toward younger age (P = 0.09; Table 1). In the University of California–San Francisco series, there was a significant correlation (P < 0.02) between baseline performance status and duration of survival.24 These findings are supported by a large multicenter study of 163 patients with HIV-related PCNSL treated with RT and evaluated for response.9 Good performance status, younger age, and higher doses of RT were correlated significantly with higher response rates,9 whereas other groups have demonstrated the prevailing significance of performance status as a predictor of outcome.38 The reported improvement in survival among patients who received ≥ 30 Gy of radiation in the current study confirms the rationale for recommending a total dose of 30–40 Gy for effective treatment of PCNSL.7–9, 22, 23 Fine and Mayer7 have recommended a shorter course of 30 Gy administered in 10 fractions for patients with poorer performance status. This dose is considered standard at most Australian and international centers.7 In the current series, an ECOG performance status of ≥ 3 was expected to be predictive of noncompletion of RT as well as poor survival for patients not receiving RT, but this was not the case. The presence of diffuse neurologic symptoms and low CD4 lymphocyte counts remained significant predictors of outcome independent of any RT effect. Nonetheless, the treatment benefit of RT persisted when these two factors were controlled for on multivariate analysis. Our analyses suggest that the patients most likely to benefit from RT were those who had a good pretreatment performance status and who were able to receive ongoing therapy with at least two antiretroviral agents. Improved survival was associated with a total RT dose of ≥ 30 Gy. Since the completion of the current study, we have routinely offered RT and HAART to patients with good performance status. In general, patients with poorer performance status are offered similar therapy, provided that our clinical impression is that they will survive beyond the planned period of RT.

Although we have demonstrated that treatment with RT, completion of RT, receipt of a total dose of 30 Gy, ongoing combination antiretroviral therapy, no more than one previous AIDS-defining illness, and the absence of nonfocal CNS symptoms and signs were predictors of prolonged survival, we have also confirmed the poor prognosis of patients with HIV-related PCNSL. The data presented in the current study suggest that RT and HAART represent the standard of care for patients with good performance. This treatment combination may serve as a baseline upon which this group of patients can be offered adjunctive novel therapies, including chemotherapy,39–41 in large multicenter clinical trials. The development of trial strategies for patients with poor performance status or for patients who do not have access to or cannot receive combination antiretroviral therapy presents an even greater challenge.

Ancillary